SERIES
PAPER
DISCUSSION
IZA DP No. 5294
The Weight of the Crisis:
Evidence from Newborns in Argentina
Carlos Bozzoli
Climent Quintana-Domeque
October 2010
Forschungsinstitut
zur Zukunft der Arbeit
Institute for the Study
of Labor
The Weight of the Crisis:
Evidence from Newborns in Argentina
Carlos Bozzoli
DIW Berlin
Climent Quintana-Domeque
Universitat d’Alacant
and IZA
Discussion Paper No. 5294
October 2010
IZA
P.O. Box 7240
53072 Bonn
Germany
Phone: +49-228-3894-0
Fax: +49-228-3894-180
E-mail: [email protected]
Any opinions expressed here are those of the author(s) and not those of IZA. Research published in
this series may include views on policy, but the institute itself takes no institutional policy positions.
The Institute for the Study of Labor (IZA) in Bonn is a local and virtual international research center
and a place of communication between science, politics and business. IZA is an independent nonprofit
organization supported by Deutsche Post Foundation. The center is associated with the University of
Bonn and offers a stimulating research environment through its international network, workshops and
conferences, data service, project support, research visits and doctoral program. IZA engages in (i)
original and internationally competitive research in all fields of labor economics, (ii) development of
policy concepts, and (iii) dissemination of research results and concepts to the interested public.
IZA Discussion Papers often represent preliminary work and are circulated to encourage discussion.
Citation of such a paper should account for its provisional character. A revised version may be
available directly from the author.
IZA Discussion Paper No. 5294
October 2010
ABSTRACT
The Weight of the Crisis:
Evidence from Newborns in Argentina*
Argentina hit headlines around the world in 2002 on account of the largest debt default in
history and a sudden economic collapse that generated statistics reminiscent of those from
the Great Depression. In this article we focus on other consequences of the crisis that are not
so obvious but that may linger for decades. Combining macroeconomic indicators with the
Argentine national registry of live births (approximately 1.9 million from 2001 through 2003),
we show that the crisis led to an average birth-weight loss of 30 grams. Our estimate is
robust to different identification strategies. This deterioration in birth weight occurred in only
about six months, and represents one-sixth of the difference in average birth weight between
American and Pakistani babies. We also find that the crisis affected particularly the weight of
babies born of mothers of low socioeconomic status. In an attempt to estimate the long-term
economic costs of the crisis, we simulate the average loss of future individual earnings due to
the reduction in average birth weight: about $500 per live birth.
JEL Classification:
Keywords:
I1, J1
Argentina, birth weight, economic crisis
Corresponding author:
Climent Quintana-Domeque
Departament de Fonaments de l’Anàlisi Econòmica
Universitat d’Alacant
Campus de Sant Vicent
Alacant 03690
Spain
E-mail: [email protected]
*
We thank Claudio Campanale, Matteo Cervellati, Vadym Lepetyuk, Marco Gonzalez-Navarro, Sonia
Oreffice, participants at the 7th Health Economics World Congress – IHEA (Beijing, July 2009), at the
FBBVA-IVIE Workshop “Health and Macroeconomics” (Madrid, December 2009), and at the 25th
European Economic Association Meetings (Glasgow, August 2010) for helpful comments and
suggestions. We thank the CIF of Universidad Torcuato Di Tella, and especially Guido Sandleris,
Ernesto Schargrodsky, and Julieta Serna for providing us access to disaggregate consumer
confidence indicators. The usual disclaimers apply. Quintana-Domeque acknowledges financial
support from the Spanish Ministry of Science and Innovation (ECO 2008-05721/ECON).
1
Introduction
Economic crises may a¤ect children’s health, and in particular their birth weight. During
recessions households may be prompted to reduce spending on items vital to children’s health,
including nutritious food and medical care for mothers and infants. Moreover, economic
downturns are likely to worsen prenatal stress, increasing the risk of adverse birth outcomes,
and may also cause public-health services to deteriorate.
We investigate the impact of the Argentine crisis that began in August 2001 on birth
weight. Argentina was shaken by a traumatic …nancial crisis at the turn of the century; its
output declined not only by about 11% in 2002 but by an additional 7% between 1999 and
2002. At the peak of the crisis, one out of four Argentines could not even a¤ord to buy basic
foodstu¤s,1 and nearly two out of three were categorized as poor.2
The occurrence of this Argentine macroeconomic episode, combined with the existence
of a national registry of live births, o¤ers the possibility of studying the e¤ect of such a
crisis on the weight of the newborns by means of data on approximately 1.9 million live
births that occurred over a three-year period, from 2001 through 2003. We …nd that in only
about six months the weight of newborns in this middle-high-income country deteriorated
by a magnitude comparable to one-sixth of the di¤erence between the average weights of
American and Pakistani newborns.
We choose birth weight as our outcome of interest for two reasons.3 First, birth weight
1
Technically, these were individuals who lived in households whose total income was below a basicfoodstu¤s basket (canasta básica alimentaria) that covers the minimal nutritional requirements for an
individual of a certain sex and age. For instance, in September 2001 the cost of the basic-foodstu¤s
basket was estimated to be $61.02 per month per adult equivalent (the exchange rate used for the
conversion was the 1 to 1 parity to the U.S. dollar). Further information can be found in an online report prepared by Argentina’s National Institute of Statistics and Censuses (INDEC), available at
http://www.indec.mecon.ar/nuevaweb/cuadros/74/pobreza2.pdf. Related statistics, derived from the periodical National Household Survey (Encuesta Nacional de Hogares), can be obtained at the INDEC website,
http://www.indec.mecon.ar
2
That is, living in households under the poverty line. For reference, the poverty line in September 2001
was estimated to be at $150.11 per month per adult equivalent.
3
Margerison (2010) o¤ers a detailed account of possible pathways linking economic recessions and birth
outcomes.
1
is a reliable predictor of survival (Mc Cormick, 1985): infants who weigh less than 2,500
grams (low-birth-weight infants) are 20 times more likely to die than are heavier babies
(UNICEF/WHO, 2004). In fact, birth-weight-speci…c neonatal mortality follows a reverse
J-pattern, with minimal risk of mortality at about 3,500 grams (Wilcox, 2001). Lower-birthweight babies have worse outcomes in terms of one-year mortality rates (Van den Berg, Lindeboom, and Portrait, 2006) and are signi…cantly shorter through childhood (Case and Paxson,
2010).4 Economic costs due to low birth weight are also substantial. For example, Almond,
Chay, and Lee (2005) show that increasing the birth weight from 2,000 to 2,500 grams reduces
inpatient hospital bills by about $10,000.
Second, recent longitudinal studies have shown that lower-birth-weight babies have worse
outcomes in terms of educational attainment, employment, and earnings (Behrman and
Rosenzweig, 2004; Case, Fertig, and Paxson, 2005; Black, Devereux, and Salvanes, 2007;
Oreopoulos, Stabile, Walld, and Roos, 2008; Royer, 2009). Moreover, researchers have also
found that the educational level attained by cohorts a¤ected by health insults in utero is
on average lower than that of adjacent birth cohorts who did not su¤er the same health insult. Almond (2006) documents the fact that children born to women infected with in‡uenza
during the 1918 pandemic received on average …ve months less education than did children
born before or conceived after this event. If economic crises a¤ect the weight of newborns,
economic recessions could result in poverty traps that a¤ect not only present welfare levels
but future ones as well.
Previous research on the e¤ects of economic crises on children’s health provides mixed
results. Paxson and Schady (2005) …nd that in Peru infant mortality increased by 2.5 percentage points during the macroeconomic crisis of the late 1980s. The authors attribute this
increase to a collapse in public and private health expenditures. In contrast, Rucci (2004)
…nds no change in Argentina’s infant-mortality rate during the crisis of the late 1990s, and
4
Using data from the NLSY Child and Young Adult Survey, Case, and Paxson (2010) …nd that children
who were heavier and taller at birth are signi…cantly taller through childhood and that taller children are less
likely to be reported to be a- icted with an emotional or neurological limiting condition.
2
credits this stability to the fact that public health expenditures were maintained at their
previous levels throughout the crisis years.
In this context, one of the main contributions of our work is the combination of a twomillion-observation sample with an actual “event” study to show that an economic collapse
can lead to large losses in birth weight in a short period of time, complementing recent work
on co¤ee-price ‡uctuations and child-survival rates in Colombia (Miller and Urdinola, 2010)
and on the role of poor conditions in the fetal period (Bhalotra, 2010).
Our work is related to a large economic-research agenda on the e¤ects of in utero insults
to birth weight, mainly nutritional de…ciencies (Ceesay et al., 1997) and psychosocial stressors
(Hobel and Culhane, 2003), at di¤erent trimesters of pregnancy.5 Economic crises can lead to
a loss of resources (worsened nutrition in utero) and psychosocial stress (Margerison, 2010).
Evidence coming from extreme cases of nutritional deprivation, such as the Dutch “Hunger
Winter” (Stein et al., 2004), of 1944-1945, during which food rations were reduced to below
1,000 Kcal per person for seven months, indicates that the birth weight of those exposed to
famine in the third trimester dropped by about 300 grams. In a recent study based on four
million childbirth events in Colombia, Camacho (2008) …nds that the intensity of random
land-mine explosions during a woman’s …rst trimester of pregnancy has a signi…cant negative
impact on her child’s birth weight. We show that birth-weight de…cits in Argentina are
correlated with economic conditions in both the …rst and the third trimesters of pregnancy
during 2001-2003.
We also …nd that socioeconomic status (proxied by mother’s education) could mitigate the
e¤ects of economic crisis, consistent with existing evidence linking maternal education and
birth-weight outcomes (e.g., Stars…eld, 1991; Currie and Moretti, 2003; Currie, 2009). The
5
Although the biological pathways linking psychosocial stressors and birth outcomes have not been completely elucidated, a neuropeptide (corticotrophin-releasing hormone, or CRH) involved in stress response and
a¤ecting the initiation of labor is thought to be a central factor. The role of CRH con…rms the …nding that
job status (in particular, jobs that demand physical exertion, and thus create physiological stress) are known
to have an adverse e¤ect on birth outcomes. Aizer, Stroud and Buka (2009) …nd that in utero exposure to elevated levels of the stress hormone cortisol negatively a¤ects the cognition, health, and educational attainment
of o¤spring.
3
average weight loss for newborns whose mothers’educational level is low is approximately 33
grams, while it is 19 grams for babies born of highly educated mothers.
Our main results are surprisingly robust to the addition of province-speci…c time trends,
suggesting that the drop in average birth weight was not driven by a reduction in either
the quality or the quantity of public resources allocated to children’s health across di¤erent
provinces over time. Additionally, if we link birth-weight statistics with regional time-varying
household-survey data, we document strong links between average birth weight and indicators
of deprivation, in particular regional extreme-poverty (indigence) rates.
In a back-of-the-envelope calculation, we estimate that the reduction in birth weight that
occurred during the economic collapse reduced the income prospects of the crisis cohort by
about $500 per childbirth. This is a lower-bound estimate that does not take into account
other long-term costs stemming from the crisis, such as the increased health burden in adult
life.6
The paper is organized as follows. In Section 2 we present a description of the data. In
Section 3 we report the estimates of the e¤ect of the Argentine crisis on average birth weight.
In Section 4 we provide an estimate of the loss in future earnings of the newborns a¤ected by
the crisis. In Section 5 we present some robustness checks. In Section 6 we conclude.
6
Adverse conditions at time of birth have been linked to heart disease, diabetes, and obesity in adulthood,
all of which contribute to a reduction in life expectancy.
4
2
The data
The main source of data for this study is the Argentine national registry of live births, Informe Estadístico del Nacido Vivo (IENV). The main strength of this dataset is its universal
coverage of all live births occurring in the country. The IENV contains information on birth
weight and weeks of gestation, but not on other child health indicators (such as AGPAR score
or head circumference). Regarding mother characteristics, there is information on her age,
parity history, marital status, and educational attainment, but not on risky behaviors such
as smoking or drinking. By de…nition, the IENV only contains information on live births,
mortality cannot be examined. This micro-level dataset contains information on 1.9 million
births occurring 2001 through 2003 in Argentina.7 Following previous work on the determinants of birth weight, we focus on mothers aged 15-49, we exclude multiple births and those
newborns whose weight was under 500 grams. Our sample size is 1,829,104 observations and
birth weight is available for 99% of births.
The core of our empirical analysis contains two well-di¤erentiated parts. We …rst use
month-by-month average-birth-weight comparisons between 2001 and 2002 to assess the e¤ect
of the crisis. In the second part, we link the state of the economy with average birth weight
on a monthly basis from January 2001 through December 2003. The state of the economy is
captured by means of an index of economic activity which replicates the ‡uctuations in the
gross domestic product (GDP), but at monthly frequencies.
2.1
Descriptive statistics
Argentina is an upper-middle-income country (World Bank, 2009), ranking as “high” in
UNDP’s Human Development Index (UNDP, 2009). In line with this ranking is its rel7
On account of a change in the structure of the birth-weight-report form, data prior to the year 2001
are not directly comparable. Our analysis therefore focuses on short-term ‡uctuations from 2001 through
2003. In practice, however, this does not appear to be a concern. Previous studies (e.g., Grandi and Dipierri,
2008) show that the decline experienced during the crisis was not a re‡ection of a secular trend but an acute
phenomenon, occurring in a matter of a few months.
5
atively small rate of low birth weights (live-birth babies weighing less than 2,500 grams;
UNICEF/WHO, 2004).Table 1 shows summary statistics regarding the period 2001-2003.8
Birth weight ‡uctuated between 3,263 and 3,231 grams, resulting in an average that is 100
grams below the U.S. standard (Martin et al., 2005). Consistent with this, the proportion of
low-birth-weight (< 2,500 g) singletons is between 6.5 and 7%, slightly above comparable U.S.
statistics. Infant Mortality Rate (IMR) increased only slightly during the economic crisis, but
resulted in a four-year period of stagnation after a period of two decades in which IMR were
halved. The second panel in the table shows selected economic indicators. Economic activity
declined by about 10% in 2002, resulting in an economy that was 10.6% below its long-term
trend in the year of the collapse. Unemployment rates reached 20% of the active population
and more than half of the population was poor, as a result of a combination of increased unemployment and a steep drop in real wages due to in‡ation pressures caused, in turn, by a sharp
depreciation of the national currency.9 Di¤erent indicators of public expenditures on child
health show a reduction of such expenditures, consistent with the …ndings of Cavagnero and
Bilger (2010).10 The last panel shows that the characteristics of the mothers remained stable
throughout the period: the mothers’average age when they gave birth was 27 years, 36% of
them were primiparous, 35-40% had completed high school, and 85% had a partner (involving
marriage or cohabitation). We use completion of high school as a proxy for high socioeconomic
status because income information is not included in the demographic-surveillance data.11
Figure 1 shows the evolution of the index of economic activity (which replicates GDP
‡uctuations at monthly frequencies) and the average birth weight of babies born from January
2001 through December 2003. There is a delay between the evolution of the economic crisis
8
Although the IENV has an extensive set of proxy measures for underlying risks of poor birth outcomes,
information to determine the health-insurance status of the mother and the employment status of the mother
and particularly of her partner is not perfectly reported. See Section 6.
9
By June 2002, the value of the peso relative to the US dollar was reduced to a quarter of what it had
been in December 2001.
10
These authors …nd that in terms of utilization of heath services, unemployment and job instability reduced
private- health-plan coverage, putting an additional strain on the already under…nanced public-health system.
11
Returns to schooling, in particular completion of secondary (high school) education and college education,
are large and thus represent a good proxy for their income opportunities (Savanti and Patrinos, 2005).
6
and the changes in average birth weight: although the crisis peaked in March 2002 (economic
activity had declined in a year by about 16% by that time) birth weight was at its nadir in
December 2002. This is what we expect, since birth weight is the cumulative e¤ect of di¤erent
inputs (e.g., nutrition, quality and quantity of medical checkups, maternal stress, etc.) during
the nine months that a pregnancy usually takes, and not just those prevailing at time of birth.
7
3
The e¤ect of the Argentine economic crisis on average
birth weight
3.1
Month-by-month estimates
In order to identify the e¤ect of the Argentine crisis on the weight of the newborns, we need to
take into account that fertility decisions are likely to be a¤ected by economic conditions, a fact
already acknowledged in the theoretical work of Becker (1991) and proven empirically both
by Dehejia and Lleras-Muney (2004), by means of US data, and, more recently, by Neugart
and Ohlsson (2009) in a quasi experiment that exploits the German parental-bene…t reform
of 2007.
As long as fertility decisions are sensitive to macroeconomic conditions, sensible methods
of estimating the e¤ect of the crisis involve accounting for the full set of characteristics of
the mother (related to both macroeconomic conditions and birth weight) or using a quasi
experiment to estimate the impact of the crisis as the mean di¤erence in birth weight between
those babies who were exposed to the crisis without their mothers’anticipating it and those
babies who were not. Otherwise, the estimated e¤ect of the crisis can be confounded by the
compositional change in the pool of childbearing women.12
There is a plethora of studies documenting the roles of di¤erent mother-and-pregnancy
characteristics relative to birth weight. In a frequently cited meta-analysis assessment, Kramer
(1987) cited 43 potential determinants of low birth weight (< 2,500 g). The most important
factors are considered to be: the age of the mother (with mothers under 20 and over 35 running
a relatively high risk of delivering a child of low birth weight); the mother’s education, which
proves to be correlated with lower rates of low birth weight (Star…eld et al., 1991); parity
and birth order (Pu¤er and Serrano, 1975); behavioral factors such as smoking and drinking,
12
However, it must be noted that even when the full set of characteristics is available, compositional changes
can create problems if there are interactions and other sources of non-linearities.
8
which have a negative impact on birth weight (Brooke et al., 1989; Cornelius et al., 1995).
We have chosen to use a mixed approach, not only accounting for characteristics of the
mother and her child but also comparing month-by-month average weights of babies born
in 2001 and 2002 whose mothers did not anticipate the crisis around the time when they
conceived.
In order to implement the quasi-experimental design, it is crucial that one …nd a cohort
of newborns who were conceived during a period when the extent of the crisis was not yet
anticipated. This assumption is plausible in light of two pieces of evidence. The …rst one
found in Figure 2, from Kannan and Köhler-Geib (2009), shows the degree of uncertainty
around the period of the Argentine crisis, measured by the dispersion of GDP forecasts based
on surveys performed by private-sector analysts: the degree of uncertainty jumped in August
2001. As long as anticipation can be proxied by lack of uncertainty, this …gure suggests that
until August 2001 the crisis was unanticipated.
The second piece of evidence supporting this assumption comes from the evolution of
the Consumer Con…dence Index for Argentina, as depicted in Figure 3, which indicates a
similar pattern in terms of expectations, with consumer-con…dence levels dropping sharply
after August 2001. Perhaps more interesting (although not reported here) is the fact that
this drop is of the same magnitude whether the consumers in question are of low or of high
socioeconomic status.13 The point is that the magnitude of the crisis was unexpected, even
though mildly pessimistic expectations may have prevailed throughout the period.
Since this paper is concerned with the impact of the crisis on live births, abortion data must
be considered. Unfortunately, not only are such data scant but the entire issue is complicated
by the fact that in Argentina abortion is illegal. A recent study by Mario and Pantelides (2009)
estimates the number of annual abortions by means of various indirect methods, adequate for
13
Since 1998 the Consumer Con…dence Index has been updated monthly by the Universidad Torcuato Di
Tella. The index is based on a monthly survey of consumer expectations similar to surveys used in OECD
countries. We thank the Center for Research in Finance (CIF) of Universidad Torcuato Di Tella, and especially
Guido Sandleris, Ernesto Schargrodsky, and Julieta Serna, for providing us with the access that we needed in
order to disaggregate consumer-con…dence indicators.
9
describing general trends but not for projecting the evolution of abortion cases from year to
year. Very crude and indirect indicators of abortion prevalence are the number of maternal
deaths due to pregnancy terminating in abortion and the number of fetal deaths. These
indicators have many shortcomings, and no discernible trend can be established by means
of data from the O¢ cial Statistical Yearbooks (Ministerio de Salud 2000-2007, Estadísticas
Vitales). Although we cannot study directly the evolution of abortion during the period
under analysis, we can proceed indirectly by looking at the ‡uctuations in the number of live
births. Figure 4 shows that this number ‡uctuates erratically, whereas average birth weight
‡uctuates systematically, relative to the economic cycle.
Our comparison group comprises babies who were both conceived and born before August
2001, while our treatment group comprises babies who were conceived before August 2001
but born after August 2001 (Fig. 5). For example, a baby conceived in July 2001, after a
normal, nine-month, gestation period, be delivered in April 2002.
In order to account for seasonality patterns in birth weight, we compare the monthly
average birth weights for January through April 2001 with those for the same four-month
period in 2002. Means of birth weight by month are estimated as the coe¢ cients of the
following model:
BWi;r;m;t =
12
X
m=1
m Im +
12
X
m Yt Im
+
r
+ Xi + "i;r;m;t
(1)
m=1
where BWi;r;m;t is the birth weight of child i born in province r in month m in year t,
Im = 1 if the month of birth is m, Yt = 1 if the year of birth is 2002,
r
= 1 if the province
of birth is r, Xi is a vector of mother-pregnancy characteristics (mother’s age, number of
pregnancies, mother’s education, and mother’s partner status), and "i;r;m;t is a random error
term.
m
is average birth weight in month m, while
m
is the di¤erence in average birth
weight in month m between 2001 and 2002. Equation (1) is estimated by OLS using clustered
standard errors at the month-by-year level (24 clusters).
10
Table 2 displays the monthly mean birth weight in 2001 and 2002 and its di¤erence. The
…rst panel uses as controls province and child-gender dummy variables. In all four month-pair
comparisons, birth weight in 2002 is lower than in 2001. The largest gap, nearly 30 grams, is
found in April, while the smallest one, about 7 grams, is found in February. Interestingly, the
30-gram gap coincides (or nearly so) with the nadir of economic activity and the peak of social
unrest. Similar estimates are reported in panel II (which adds mother’s age and pregnancy
categories as controls) and panel III (which controls for the mother’s education and partner
status as well).
These …ndings suggest that the crisis had a negative e¤ect on birth weight, which was
about 30 grams for babies born in the month of April 2002. Babies born earlier in the crisis
were less a¤ected than those born in February and March, probably because the e¤ects of
the crisis operate in utero in a cumulative manner: children born in April 2002 were exposed
to nine months of the crisis, while those born in February and March 2002 were exposed to
seven and eight months of the crisis, respectively.
Table 3 disaggregates results by the sex of the newborn. Although all comparisons point
toward a reduction in birth weight, boys are slightly more a¤ected than girls. These …ndings are consistent with some evidence that boys are particularly vulnerable to food supply
shortages in utero (Eriksson, Kajantie, Osmond, Thomburg, and Barker, 2010).
In Table 4 we inquire about the di¤erential e¤ect of the crisis on average birth weight
depending on the mother’s socioeconomic status, proxied for whether the mother completed
completed high school. Regardless of the set of controls used, the decline in birth weight is
particularly prevalent in boys and girls born to low-socioeconomic status mothers. If anything,
a higher socioeconomic status appears to cushion newborns during an economic crisis.
11
3.2
3.2.1
Economic cycle and average birth weight
Economic cycle at birth and average birth weight
To assess the state of the economy, we calculate the deviation of the economic-activity indicator with respect to its long-term trend (expressed in log units). This deviation is usually
referred to as the cyclical component, in that it isolates business-cycle ‡uctuations. We use a
Hodrick-Prescott …lter, which is a standard decomposition method of identifying ‡uctuations
at business-cycle frequencies (i.e., booms and recessions).14 In the case under consideration,
the economy plunges into a recession so quickly that by mid-2002 economic activity is more
than 10% below its long-term trend.
We estimate models of the form
BWi;r;m;t = Cm + Im +
r
+ Yt + Xi + "i;r;m;t
(2)
BWi;r;m;t = Cm + Im +
r
+ t + Xi + "i;r;m;t
(3)
where Cm is the cyclical component of the economic-activity indicator during the month
of birth m. Both models (2) and (3) contain two types of control variables. The …rst set
of controls includes: month of birth-…xed e¤ects Im , to account for seasonality patterns in
birth weight; province of birth-…xed e¤ects
r,
to capture regional di¤erences in the health-
care infrastructure and other factors that are …xed in time but vary across provinces; and
time e¤ects (either year-…xed e¤ects Yt in (2), or a linear time trend t in (3)) to account for
secular trends in birth weight. However, given our small time window (2001-2003), using a
time trend, comes at a risk: the overestimation of the secular decline in birth weight.15 The
14
Since we are using monthly data, we choose a smoothing parameter of 129,600 (Ravn and Uhlig, 2002).
Our …ndings are not sensitive to the method used, as our use of other …ltering methods attests.
15
Grandi and Dipierri (2008) use data from 1992 through 2002 and …nd a secular decline of about 2 grams
per year in the birth weight of Argentine babies. The decline in birth weight that occurred in 2002 alone
amounts to 30 grams (Table 1). Moreover, the secular trend in the 2001-2003 sample may absorb business-
12
second set of controls, Xi , includes: mother’s-age categories, parity categories, an indicator of
whether the mother has completed high school, an indicator of whether the mother is living
with his partner (married or cohabiting), and the interaction of these last two variables.
Unfortunately, information on mothers’smoking and drinking habits/patterns is not included
in the birth-registry.
Table 5 displays a series of regressions of birth weight on economic cycle at birth and
other variables for the full sample of boys and girls. The estimates indicate that average
birth weight is positively associated with the cycle at birth. In other words, average birth
weight is a procyclical variable. It should be noted, however, that if in utero conditions a¤ect
birth outcomes in a cumulative manner (that is, if they operate, to a greater or lesser degree,
throughout the entire gestation period), one would expect business-cycle lags, too, to have an
in‡uence on birth weight.
Looking at the rest of the coe¢ cients on the table we can see that at birth girls are on
average 103 grams lighter than boys: a …nding similar to one reported in Kramer (1987).
Newborns of highly educated mothers are heavier than those whose mothers are not, a …nding that resonates with others linking maternal education and birth-weight outcomes (e.g.,
Stars…eld, 1991; Currie and Moretti, 2003; Currie, 2009).
In Table 6 we report estimates from the last two speci…cations of Table 5 for boys and
girls separately. The estimated impact of the business cycle is slightly more important for
boys than for girls: a …nding consistent with the estimates presented in Table 3, above.
Table 7 presents evidence on the di¤erential impact of the crisis on the weight of newborns
according to their mothers’educational level. The table shows that in every speci…cation the
business cycle at birth is strongly correlated with the birth weight of children born to mothers
whose educational level is low, and that the impact on children born to highly educated
mothers at least doubles the e¤ect on the others. In fact, a speci…cation using a linear
cycle ‡uctuations and could theoretically reduce the estimates of the impact of business-cycle ‡uctuations
(i.e. the collapse) on birth weight for this particular sample period.
13
time trend indicates that the e¤ect of the business cycle during the month of birth for this
subsample of children is insigni…cant.16
Mothers of low socioeconomic status had on average lighter babies than did the others
(Tables 5 and 6). Moreover, less-educated mothers were hit harder by the crisis (Table 7).
In other words, babies born into poor families have a disadvantage in normal times (without
recessions) which becomes even wider in bad times (with recessions).
What would be the observed evolution of average birth weight had the economy not entered
in the recession? Figure 6 shows the evolution of actual average birth weight, what the average
birth weight would have been – according to the model in column (6) of Table 5 – had the
economy not entered in the recession, and the di¤erence between them.17 This simulation
(hypothetical situation) is achieved by replacing the actual monthly cyclical component from
August 2001 to the end of our time window with randomly chosen realizations of the cyclical
components from January 2001 through July 2001. The negative e¤ect of the crisis in terms
of birth-weight peaks during the third quarter of 2002, six months after the business cycle
trough (see Fig. 1) and its adverse in‡uence declines over time on account of the economic
recovery that ensued.
3.2.2
Economic cycle during the trimesters of pregnancy and average birth
weight
Since birth weight is a¤ected by economic conditions throughout pregnancy, attempting to
capture the impact of the economic crisis by using only the economic cycle at birth may not
be accurate. Thus, for each birth we create a measure of the economic cycle in each of the
three quarters that a pregnancy usually takes. For the …rst quarter of pregnancy, we take
16
If in addition we break down the sample by gender of the child, reestimation of column (3) separately for
boys and girls gives us cycle coe¢ cients of 147.96*** (27.51), and 136.62*** (32.37), respectively. Reestimating
column (6), we obtain coe¢ cients of 28.57 (31.48) and 21.27 (23.59) for boys and girls, respectively. Similar
qualitative results are obtained if instead of using a linear time trend we control for year-…xed e¤ects. More
speci…cally, the signi…cant estimates, which are obtained for those childbirths involving mothers with a low
educational level, are 120.53** (49.53) and 100.04** (42.76) for boys and girls, respectively.
17
Note that the period during which the comparison takes place now begins with August 2001.
14
the average of the monthly cyclical component in those three initial months, C1 , and we do a
similar procedure for the second and third quarters of pregnancy, C2 and C3 .
We start by estimating models in which we include the cyclical component for only one of
the trimesters
BWi;r;m;t =
where
T
T CT
+ Im +
r
+ T ime + Xi + "i;r;m;t
(4)
re‡ects the sensitivity of birth weight to economic conditions during trimester T
of pregnancy, and T ime is either a set of year-…xed e¤ects, Yt , or a linear time trend, t. Table
8 shows that di¤erent indicators of the business cycle during pregnancy have e¤ects of the
same order of magnitude on birth weight. This is not surprising given the high autocorrelation
displayed by business-cycle deviations and the fact that indicators are introduced separately.18
The impact of quarterly indicators (cycle in a given trimester of pregnancy) shown in Table
8 is much greater than the e¤ect of the business cycle during the month of birth. Although
speculative, this result is expected if monthly time series are noisier than quarterly averages,
or if the e¤ect of adverse economic conditions is cumulative throughout the pregnancy period.
The next step is to include the cycles in each trimester simultaneously, in a single equation:
BWi;r;m;t =
3
X
T CT
+ Im +
r
+ T ime + Xi + "i;r;m;t
(5)
T =1
Table 9 shows that economic conditions during the …rst and third quarters signi…cantly a¤ect
birth weight. The cumulative impact of the three trimesters is comparable to the estimates
shown in Table 8. Using these estimates, a deviation of 0.1 log units (about 11%) from the
long-term trend (similar to that observed in 2002, as shown in Table 1) would explain a
reduction in birth weight of about 25-30 grams. Using the results in Table 9, we can predict
the average birth weight lost on account of business-cycle ‡uctuations in di¤erent months.
18
The correlation between the cyclical components is: 0.8866 between the third and second trimesters of
pregnancy; 0.6437 between the third and the …rst; and 0.9000 between the second and the …rst.
15
For example, using estimates in model (5) of Table 9 as a reference, we predict average
birth weight under the actual scenario and under a counterfactual (no crisis) situation that
assumes macroeconomic conditions in those observed in August 2001 and other regressors
…xed at their observed levels. The di¤erence between the predicted and the counterfactual
outcomes is averaged by monthly cohorts (by month of birth) and shown in Figure 7, along
with the monthly business-cycle indicator.
Table 10 shows how the results in columns (5) and (6) of Table 9 change when the sample
is broken down by gender: the e¤ect on boys is somewhat greater than it is on girls. When the
sample is split according to the mother’s educational level, we …nd that the sensitivity of birth
weight to economic conditions is substantially larger for children of mothers whose educational
level is low (Table 11). This subset is sensitive to economic conditions in the …rst and last
trimesters of pregnancy, while children of mothers of high socioeconomic status are sensitive
to early-pregnancy economic conditions. Thus, depending on the subsample studied, results
are consistent with the famine literature (highlighting nutritional-deprivation factors) or with
the quasi-experimental results of Camacho (2008) from Colombia (emphasizing psychosocialstress factors). Using the estimate from the sum of the cyclical coe¢ cients, a deviation of 0.1
log units (about 11%) from the long-term trend (similar to that observed in 2002, as shown
in Table 1) would explain a reduction in birth weight of about 31-35 grams for babies born of
mothers with a low educational level and 17-20 grams for babies born of mothers with a high
level.
16
4
The long-lasting e¤ect of the Argentine economic crisis: a simulation exercise
For those whose births were a¤ected by the crisis, what is its long-term impact? We o¤er a
tentative answer to this question by simulating lifetime earnings of these individuals under
di¤erent assumptions regarding their working life (the age at which they start working and
the age at which they retire) and income-growth patterns (the rate at which incomes increases
from one year to the next). We calibrate our model with Argentine data on income in current
purchasing-power-parity (PPP) dollars, as is standard in the specialized literature of crosscountry comparisons.
Our previous results suggest that the average birth-weight loss associated with the crisis
was about 30 grams. Since we are interested in the impact of a reduction in birth weight on
lifetime earnings, we compare the earnings path of an individual born during the recession
(with the 30-gram birth-weight loss) with the counterfactual income path for an individual
not born in the recession (without the 30-gram birth-weight loss). For the no-recession path
of income, we assume that individuals earn a level of income equal to the expected (national)
GDP per capita for each year they are in the labor force. Expected future income is based on
a baseline GDP per capita for 2009 and an annual-income-growth rate that varies from 1% to
5% per year. To calculate the income loss we use results from Black, Devereux, and Salvanes
(2007) to approximate
ln(Income)/ ln(Birthweight), as shown in the footnote of Table 12,
which presents our estimates.19 We …nd that the average loss of future earnings due to the
reduction in average birth weight is about $500 per baby born, although the magnitude (but
not the sign) of the costs is sensitive to key model assumptions (namely, expected income
growth and intertemporal discount rate). However, it is very likely that these costs exceed
19
The impact of birth weight on lifetime income has been estimated using within-twin variation (Behrman
and Rosenzweig, 2004; Black, Devereux, and Salvanes, 2007). One should keep in mind that although these
estimates control for unobservable factors capturing the in utero environment, they are based on a selected
sample (twins).
17
those of measures designed to prevent birth-weight loss. For example, eliminating poverty
among pregnant mothers (by raising their income to the poverty line) would cost only $100
per mother to do so for the entire nine-month period of pregnancy, which can be considered
an overestimate of the cost of preventing the drop of birth weight from occurring.
What can be done to prevent such costs? Since birth outcomes are a¤ected by di¤erent
factors, there is no simple recipe to avoid these problems. But there is ample evidence that
targeted interventions work, even amid poverty, where they are most needed (Ramakrishnan,
2004). Depending on speci…c circumstances, these interventions would include a combination
of strategies: nutritional supplementation, provision of adequate prenatal care, and promotion
of maternal behavioral changes associated with better health outcomes.20
One must keep in mind that the $500 estimate does not take into account other long-term
costs stemming from the crisis, such as increased health burdens in adult life; : adverse conditions at time of birth have been linked to heart disease, diabetes, and obesity in adulthood,
all of which contribute to a reduction in life expectancy.
20
See Almond, Hoynes and Schanzenbach (2010) for a recent study on the impact of food stamps on birth
outcomes in the US.
18
5
Robustness checks
5.1
Additional controls
We have yet to consider the in‡uence of the type of health coverage (public or private), the
location in which the birth took place (public or private hospital/clinic, home, or street), and
who aided in the delivery of the baby (doctor or someone else).
For 1,785,384 observations we have information on whether a doctor or someone else
aided in the delivery of the baby. In 2001, 72% of childbirths were attended by a doctor. The
percentages were very similar in 2002 and 2003: 71% and 72%, respectively.
Regarding the health coverage of mothers who gave birth during the period under analysis,
we have information on only 88% of childbirths. Twenty-…ve percent of them had public
health-care coverage, 2.8% had coverage with the private sector, 0.3% had both, and 38% had
none. Because our dataset is incomplete, we have a large percentage of unknowns (38%) and
a few miscodings (736 births).
Finally, our dataset also contains information on the location where the birth took place.
However, it is far from perfect. We observe the following location categories for the entire period: public hospital/clinic (35.75%), private hospital/clinic (22.5%), home or street (0.75%),
miscoded (29.15%), and unknown (11.85%).
In Table 13 we report speci…cations similar to Table 9 (columns 5 and 6), but including the
controls mentioned above. The sensitivity of birth weight to business cycle is not substantially
altered with respect to those shown in Table 9. This indicates that our …ndings are robust to
any compositional changes operating through observable health inputs.
5.2
Alternative indicators of economic activity: indigence and poverty
Our analysis of economic activity and birth weight relied on the (HP-detrended) cyclical component of economic activity at the national level and in monthly frequency. However, it is
19
possible that there was regional variation in the magnitude of the crisis and/or that it presented lag/lead shifts with respect to the national indicator. Unfortunately, no monthly (or
even quarterly) indicator of economic activity is available at the provincial level. Nevertheless,
the household survey of urban areas (Encuesta Permanente de Hogares)21 provided us with
information on poverty and indigence rates in 29 such urban conglomerates, which represent
urban populations in 22 provinces and the Federal District (Ciudad de Buenos Aires). Information on poverty indicators is available for May and October for each year since 2001. Data
after May 2003 are not comparable on account of a change in the methodology. Because the
data are collected twice a year in particular months, we have extrapolated missing observations linearly to generate monthly observations. This procedure is not fail-safe, but it bene…ts
from the fact that the periods in which the data were collected were near turning points in
the business cycle (see Fig. 1).
We calculate poverty rates at the district level in the third, second, and …rst trimesters of
pregnancy, respectively. Poverty rates are calculated on the basis of international standards
by the o¢ cial National Institute of Statistics and Censuses of Argentina. The poverty line
varies according to household composition and the month/year in which the survey is carried
out, but, for example, $149 per month per adult equivalent in December 2001, or about $5
per day, using the prevailing exchange rate at that time (not PPP).
Similarly, the indigence indicators represent the indigence rate (or extreme-poverty rate) at
di¤erent trimesters of pregnancy. The indigence rate is calculated by means of an “indigence
line”: for example, in December 2001 this indicator was $60 per month per adult equivalent,
or about $2 per day. The indigence line is the sum required to buy a basic-foodstu¤s basket (Canasta Básica Alimentaria) that meets minimum nutritional requirements, while the
poverty line is based on the indigence line plus additional expenditures on basic nonfood items
(e.g., transportation, housing, and clothing).
21
Only the urban population of one province (Rio Negro) was not included systematically from the start
(but instead from October 2002 on).
20
Table 14 displays regressions of birth weight on poverty and indigence rates. The …rst
column shows a speci…cation with moving averages of the indigence rates. Coe¢ cients are
jointly, but not individually, signi…cant. Similar results are obtained for three regressors
involving moving averages of poverty rates. We therefore proceed by eliminating one of the
regressors in each speci…cation. Column (3) shows that indigence rates at the district level
in the three last months of pregnancy signi…cantly decrease birth weight. The same result
is found in column (4) for poverty in the last three months of pregnancy, although poverty
rates in the …rst three months exhibit a positive (but insigni…cant) sign. However, the sum
of both coe¢ cients is negative, and signi…cantly so (p-value=0.045). Column (5) combines
poverty and indigence (both in the last term of pregnancy) as predictors of birth weight, with
indigence (extreme poverty) proving to be signi…cant, with the expected sign.
These results show that poverty and indigence may have a deprivation e¤ect on birth
weight. These …ve models have very di¤erent predictive capabilities. Figure 8 displays the
change in birth weight using August 2001 as baseline and using the temporal variation in
regressors speci…ed in each column of Table 14, together with their estimated coe¢ cients. For
purposes of comparison we present the actual variation in birth weight and the prediction
based in the model with the national business-cycle indicator displayed in Table 9 column
(5). The models that combine time and regional variation are less powerful in predicting ‡uctuations in birth weight, with the model in column (5) being the one with the best predictive
power. This result can be interpreted in a number of ways. First, the fact that poverty and
indigence indicators are monthly extrapolations based on actual indicators calculated every
six months may weaken their predictive ability. Second, poverty and indigence indicators
may explain variation when it comes to mothers with a low educational level, whereas other
indicators may be more useful predictors when it comes to mothers with a high educational
level. Finally, the models shown in Table 14 use information from June 2001 through May
2003. The methodology of household survey changes after May 2003, no disaggregated data
21
by urban areas are available for the period prior to October 2000. The temporal dimension
of the sample, lacking as it does the early period of 2001 (before the crisis hit) and the 2003
recovery period, is thus reduced.
These estimates suggest that the results from the national monthly business-cycle indicator
may be –at least in part –capturing a deprivation e¤ect that is mediated by indigence and
poverty, which are good predictors of the ‡uctuations in birth weight during the crisis.
5.3
Additional checks: transformations of the dependent variable
and province-speci…c-time/month e¤ects
We will now highlight two of the additional checks we have performed in order to assess the
validity of our estimates. First, when we replace birth weight with its logarithm and when we
use fetal growth, de…ned as birth weight over weeks of gestation, we obtain similar qualitative
results. Second, the addition of province-speci…c linear time trends, province-speci…c month
of birth-…xed e¤ects, and even interactions between month of birth and a linear time trend
leads to estimates that are of the same order of magnitude as that of those reported in the
previous speci…cations.22
22
All of these estimates are available from the authors upon request.
22
6
Conclusions
Economic crises are episodic phenomena that raise concerns about unemployment, poverty,
and bailouts, terms all too familiar to the public at large. In this article we focus on less
obvious –but perhaps equally important –costs stemming from an economic crisis: a birthweight loss. This burden is likely to have long-lasting e¤ects, since birth weight in‡uences
lifetime earnings, as has been shown in the literature.
We …nd that the crisis explained a loss of about 30 grams in average birth weight. It is
important to keep in mind that this loss occurred in a short period of time (about six to seven
months) and that this is a population average: the impact of the recession was even more
pronounced among mothers of low socioeconomic status.
In an attempt to estimate the long-term economic costs of the crisis, we simulate the
average loss of future individual earnings due to the reduction in average birth weight: about
$500 per live birth. This is a conservative estimate because it does not include other potential
losses that are not re‡ected in lifetime earnings, such as lifetime health-care costs and a
reduction in life expectancy. The price paid will be higher for some than for others, since
low birth weight is associated with poor mothers, a discrepancy that may exacerbate income
inequalities in the long run.
Another reason that our results are striking is that such a disruption in health status
occurred in a middle-to-high-income country with a physician-to-patient ratio similar to those
of Germany and Norway. Perhaps our most important conclusion is that the adverse e¤ects
of economic crises on children health may not be restricted to times or to places distinguished
by extreme poverty.
23
References
[1] Aizer, A., Stroud, L., and S. Buka (2009) “Maternal Stress and Child Well-Being: Evidence from Siblings,”Brown University, mimeo.
[2] Almond, D., Chay, K. Y., and D. S. Lee (2005) “The Costs of Low Birth Weight,”
Quarterly Journal of Economics, 120(3):1031–1083.
[3] Almond, D. (2006) “Is the 1918 In‡uenza Pandemic Over? Long-term E¤ects of in utero
In‡uenza Exposure in the Post-1940 U.S. Population,” Journal of Political Economy,
114(4):672–712.
[4] Almond, D., Hoynes, H., and D. Schanzenbach (2010) “Inside the War on Poverty:
The Impact of Food Stamps on Birth Outcomes,” Review of Economics and Statistics,
forthcoming.
[5] Becker, G. (1991) A Treatise on the Family. Harvard University Press.
[6] Behrman, J.R. and M. Rosenzweig (2004) “Returns to Birthweight,” Review of Economics and Statistics, 86 (May):586–601.
[7] Bhalotra, S. (2010) “Fatal Fluctuations? Cyclicality In Infant Mortality In India,”Journal of Development Economics, forthcoming.
[8] Black, S. E., Devereux, P.J., and K.G. Salvanes (2007) “From the Cradle to the Labor Market? The E¤ect of Birth Weight on Adult Outcomes,” Quarterly Journal of
Economics, 122(1):409–439.
[9] Brooke, O.G., Anderson, H.R., Bland, J.M., Peacock, J.L., and C.M. Stewart (1989)
“E¤ects on birth weight of smoking, alcohol, ca¤eine, socioeconomic factors, and psychosocial stress,”British Medical Journal, 298:795–801.
24
[10] Camacho, A. (2008) “Stress and Birth Weight: Evidence from Terrorist Attacks,”American Economic Review Papers and Proceedings, 98(2):511–515.
[11] Case, A., Fertig, A., and C. Paxson (2005) “The lasting impact of Childhood Health and
Circumstance,”Journal of Health Economics, 24: 365–389.
[12] Case, A., and C. Paxson (2010) “Causes and consequences of early life health,” NBER
Working Paper # 15637.
[13] Cavagnero, E., Bilger, M. (2010) “Equity during an economic crisis: Financing of the
Argentine helath system”. Journal of Health Economics, in press.
[14] Ceesay, S., Prentics, A., Cole, T., Foord, F., Poskitt, E., Weaver, L., and R. Whitehead
(1997) “E¤ects on birth weight and perinatal mortality of maternal dietary supplements
in rural Gambia: 5 year randomised controlled trial,”British Medical Journal, 315:786–
790.
[15] Cornelius, M.D., Taylor, P.M., Geva, D., and N. Day, N. (1995) “Prenatal Tobacco and
Marijuana Use Among Adolescents: E¤ects on O¤spring Gestational Age, Growth, and
Morphology,”Pediatrics, 95(5):738–743.
[16] Currie, J., and E. Moretti (2003) “Mother’s Education and the Intergenerational Transmission of Human Capital: Evidence from College Openings,”Quarterly Journal of Economics, 118(4):1495–1532.
[17] Currie, J. (2009) “Healthy, Wealthy, and Wise: Socioeconomic Status, Poor Health in
Childhood, and Human Capital Development,” Journal of Economic Literature, 47(1):
87–122.
[18] Dehejia, R. and A. Lleras-Muney (2004) “Booms, Busts and Babies’Health,”Quarterly
Journal of Economics, 119(3):1091–1130.
25
[19] Eriksson, J. G., Kajantie, E., Osmond, C., Thornburg, K. and D. J. Barker (2010) “Boys
live dangerously in the womb,”American Journal of Human Biology, 22:330–335.
[20] Grandi, C. and J. Dipierri (2008) “Tendencia secular del peso de nacimiento en Argentina
(1992-2002): un estudio poblacional,”Archivos Argentinos de Pediatría, 106(3):219–225.
[21] Hobel, C. and J. Culhane (2003) “Role of Psychosocial and Nutritional Stress on poor
Pregnancy Outcomes,”Journal of Nutrition, 133:1709S–1717S
[22] Homer C., James S., and E. Siegel (1990) “Work-related psychosocial stress and risk of
preterm, low birthweight delivery,”American Journal of Public Health, 80(2):173–177.
[23] Kannan, P. and F. Köhler-Geib (2009) “The uncertainty channel of contagion,” The
World Bank, Policy Resarch Working Paper 4995.
[24] Kramer, M.S. (1987) “Determinants of low birth weight: methodological assessment and
meta-analysis,”Bulletin World Health Organization, 65(5):663–737.
[25] Margerison Zilkol, C.E. (2010) “Economic contraction and birth outcomes: an integrative
review,”Human Reproductive Update, 16(4):445-458.
[26] Mario, S and Pantelides, E. (2009) “Estimación de la magnitud del aborto inducido en
Argentina”. Notas de Población CEPAL 87:85–120.
[27] Martin, J.A., Hamilton, B.E., Sutton, P.D., Ventura, S.J., Menacker, F., and M. L.
Munson (2005) “Births: Final Data for 2003,”National Vital Statistics Reports, Volume
54, Number 2 September 8, 2005.
[28] McCormick, M.C. (1985) “The contribution of low birth weight to infant mortality and
childhood morbidity,”New England Journal of Medicine, 312:82–90.
26
[29] Miller, D., and B. Piedad Urdinola (2010) “Cyclicality, Mortality, and the Value of
Time: The Case of Co¤ee Price Fluctuations and Child Survival in Colombia,” Journal
of Political Economy, 118 (1):113–155.
[30] Ministerio de Salud de la Republica Argentina, various editions (2000-2007), Estadisticas
Vitales: Informacion Basica. Buenos Aires: Direccion de Estadisticas en Informacion en
Salud (DEIS). Ministerio de Salud.
[31] Neugart, M., and H. Ohlsson (2009) “Economic incentives and the timing of births:
evidence from the German parental bene…t reform 2007,” Uppsala Center for Fiscal
Studies, Department of Economics, Working Paper 2009:10.
[32] Oreopoulos, P., Stabile, M., Wald, R., and L.L. Roos (2008) “Short-, Medium-, and
Long-Term Consequences of Poor Infant Health: An Analysis Using Siblings and Twins,”
Journal of Human Resources, 43(1):88–138.
[33] Paxson, C., and N. Schady (2005) “Child Health and Economic Crisis in Peru,” World
Bank Economic Review, 19(2):203–223.
[34] Pu¤er, R.R. and C.V. Serrano (1975) “Birthweight, maternal age, and birth order: three
important determinants in infant mortality,” Washington, D.C., Pan American Health
Organization. (Scienti…c Publication, No. 294).
[35] Ramakrishnan, Usha (2004) “Nutrition and low birth weight: from research to practice,”
American Journal of Clinical Nutrition, 79:17–21.
[36] Ravn, Morten O. and Harald Uhlig (2002) “On adjusting the Hodrick-Prescott …lter for
the frequency of observations,”Review of Economics and Statistics, 84(2): 371–376.
[37] Royer, Heather (2009) “Separated at Girth: US Twin Estimates of the E¤ects of Birth
Weight,”American Economic Journal: Applied Economics, 1(1): 49–85.
27
[38] Rucci, Graciana (2004) “The Role of Macroeconomic Crisis on Births and Infant Health:
The Argentine Case,”University of California, Los Angeles, Department of Economics.
[39] Savanti, Maria Paula and Harry Patrinos (2005) “Rising returns to schooling in Argentina, 1992-2002 : productivity or credentialism?,”Policy Research Working Paper
Series 3714, The World Bank.
[40] Star…eld, B., Shapiro, S., Weiss, J., Liang, K-Y., Ra, K., Paige, D. and X. Wang (1991)
“Race, Family Income, and Low Birth Weight,” American Journal of Epidemiology,
134(10): 1167–1174.
[41] UNDP (2009) Human Development Report 2009.
[42] UNICEF
gional
and
and
WHO
global
(2004)
“Low
estimates,”
birthweight:
UNICEF,
New
Country,
York,
re2004.
http://www.who.int/making_pregnancy_safer/documents/9280638327/en/
[43] Van den Berg, G. J., Lindeboom, M., and F. Portrait (2006) “Economic Conditions Early
in Life and Individual Mortality,”American Economic Review, 96(1):290–302.
[44] Wilcox, A. J. (2001) “On the importance-and the unimportance-of birthweight,” International Journal of Epidemiology, 30:1233–1241.
[45] World Bank (2009) World Development Report 2009.
28
2001m1
3200
3220
3240
3260
Avg. Birth Weight (g)
Economic Activity Index, 1993=100
100
105
110
115
120
3280
Fig. 1: Economic Activity Index and Average Birth Weight
January 2001 – October 2003
2001m7
2002m1
2002m7
2003m1
Year and Month of Birth
Economic Activity Index, 1993=100
2003m7
Avg. Birth Weight (g)
Fig. 2: Evolution of Uncertainty
Source: Figure 2 in Kannan and Köhler-Geib (2009).
30
Consumer Confidence Index
35
40
45
50
Fig. 3: Evolution of Consumer Confidence
2001m1
2001m7
2002m1
Year and Month
2002m7
2003m1
2001m1
46000
48000
50000
52000
Number of live births
Economic Activity Index, 1993=100
100
105
110
115
54000
120
Fig. 4: Evolution of Live Births
2001m7
2002m1
2002m7
2003m1
Year and Month of Birth
Economic Activity Index, 1993=100
2003m7
Number of live births
Fig. 5: Comparison and Treatment
Year
01 01 01 01 01 01 01 01 01 01 01 01 02 02 02 02
Month J
F M A M J
Comparison period
J
A S O N D
J
F M A
Treatment period
3260
3220
3240
Difference in BW (g)
-20
-10
3200
-30
2001m7
Avg. Birth Weight (g)
0
3280
Fig. 6: Loss in Average Birth Weight due to the Crisis
August 2001 – October 2003
2002m1
2002m7
2003m1
Year and Month of Birth
Difference in BW (g)
Avg. Birth Weight (g)
2003m7
Avg. Birth Weight (g), Hyp.
Note: Avg. Birth Weight (g), Hyp.: predicted average birth weight (in grams) using model in column (6)
of Table 5 replacing cycles from August 2001 to October 2003 with cycles from January 2001 to July
2001 // Avg. Birth Weight (g): actual average BW, in grams, shown in right scale // Difference in BW (g):
difference between them.
0
-.15
-40
-30
Grams
-20
-10
-.1
-.05
Deviation from HP Trend in log-units
'
0
Fig. 7: Average birth weight loss due to business cycle fluctuations
in different months, August 2001 – October 2003
2001m7
2002m1
2002m7
2003m1
2003m7
Year and Month of Birth
Birth weight loss
Cyclical component (right axis)
-60
-40
-20
0
20
Fig. 8: Comparisons of economic activity, poverty and indigence models
2001m4
2001m10
2002m4
2002m10
Year and Month of Birth
Predicted by Col 1
Predicted by Col 3
Predicted by Col 5
Birth Weight loss (g), Actual
2003m4
Predicted by Col 2
Predicted by Col 4
Predicted by Table 9 Col 6
Table 1: Descriptive Statistics, means and (standard deviations)
2001
2002
2003
3263.33
(543.64)
0.065
(0.246)
0.488
(0.500)
16.3
3235.93
(538.36)
0.069
(0.253)
0.486
(0.500)
16.8
3231.30
(541.07)
0.070
(0.256)
0.487
(0.500)
16.5
Economic Indicators
Economic Activity Indexb (1993 = 100)
Economic Cyclec
Unemploymentd
Povertye
95.73
0.0046
17.4%
37.1%
85.30
−0.1059
19.7%
55.3%
92.84
−0.0433
15.6%
54.7%
Public Expenditure in Children Health (2006 PPP $)f
National Budget per childg
National Budget on Mother-Infant Program per child
National and Provincial Budget per childg
24.39
4.17
268.61
30.72
5.63
192.68
31.14
8.11
189.34
26.62
(6.44)
0.361
(0.480)
0.356
(0.479)
0.853
(0.354)
26.59
(6.44)
0.353
(0.478)
0.370
(0.483)
0.834
(0.372)
26.94
(6.40)
0.361
(0.480)
0.395
(0.489)
0.848
(0.359)
Birth Outcomes
Birth Weight (g)
Low Birth Weight
Female
Infant Mortality Ratea
Characteristics of the mother
Age (years)
First pregnancy
High School
Partner (married or cohabiting)
Note: Number of observations to calculate live birth characteristics (birth weight, low birth weight, female, age of the mother,
first pregnancy, mother has high-school or above, and mother has a partner) are 595,980 in 2001, 581,188 in 2002, and 548,257
in 2003. The number of observations in 2003 is “artificially” smaller than in 2001 and 2002, since childbirths occurring in the last
three months of the year are statistically reported with a lag, and our dataset does not capture the updates occurring after 2003.
b
a
Source: Argentine Ministry of Health, Yearbook. // Indicador Sintetico de la Actividad Economica. Source: Instituto Nacional
de Estadisticas y Censos (INDEC) // c Cyclical component of log Economic Activity Index (in log units) // d National Average of
Unemployment Rate for May/October (2001 and 2002), and May (2003). Source: Instituto Nacional de Estadisticas y Censos
e
(INDEC) // Proportion of Individuals under the Official Poverty Line. National Average for /October (2001 and 2002), and May
(2003). Source: Instituto Nacional de Estadisticas y Censos (INDEC) // f Source: DAGPyPS/Unicef (2007) “Gasto Publico Social
dirigido a la Niñez en la Argentina 1995-2007” Available online: http://www.gastopubliconinez.gov.ar/inversion_n_04.php.
Nominal values are converted to 2006 pesos using a mixed CPI-WPI price index and then converting to PPP dollars at the parity
of 2006. // g Includes mother-infant programs, prevention programs, vaccination, school health, medication, outpatient/inpatient
services, organ transplantation, sexual/reproductive health, AIDS/HIV and other STDs and other services and goods provided by
central and provincial government and targeted to individuals ages 0-17.
Table 2: Differences in average birth weight (g) between 2002 and 2001
2001
2002
I. Controls: province and child gender dummy variables
January
3413.77
3400.82
February
3420.82
3413.38
March
3414.10
3395.93
April
3420.59
3390.87
N
Difference
−12.95***
(0.043)
−7.44***
(0.052)
−18.17***
(0.048)
−29.72***
(0.055)
1,238,320
II. Controls: I + age and pregnancy categories
January
3370.96
3357.33
February
3377.89
3370.49
March
3370.93
3353.52
April
3378.31
3349.06
N
−13.63***
(0.050)
−7.40***
(0.049)
−17.41***
(0.056)
−29.25***
(0.055)
1,223,823
III. Controls: II + mother’s education and partner dummy variables
January
3324.22
3311.45
February
3330.97
3324.15
March
3323.78
3308.43
April
3331.33
3304.10
N
−12.77***
(0.188)
−6.82***
(0.179)
−15.35***
(0.179)
−27.22***
(0.183)
1,153,457
Note: OLS regressions of birth weight on month of birth indicators, their interactions with 2002, and
controls. Robust standard errors clustered at the “month-year” of birth level are reported in parentheses.
*** p-value < 0.01, ** p-value < 0.05, * p-value < 0.1
I: excluded province (jurisdiction) is “Tierra del Fuego”
II: excluded age category is “45-49” and excluded pregnancy category is “4 or more”
III: mother’s education dummy variable is 1 if high-school or above, 0 otherwise; partner dummy
variable is 1 if living with a partner, 0 otherwise.
Table 3: Differences in average birth weight (g) between 2002 and 2001 by gender
Boys
Girls
2001
2002
difference
2001
2002
I. Controls: province and child gender dummy variables
January
3414.16
3399.34
February
3424.59
3413.92
March
3417.40
3395.60
April
3420.80
3390.20
N
−14.82***
(0.051)
−10.67***
(0.078)
−21.80***
(0.068)
−30.60***
(0.077)
635,151
3310.37
3299.32
3313.87
3309.84
3307.65
3293.32
3317.40
3288.59
Difference
−11.05***
(0.067)
−4.03***
(0.067)
−14.33***
(0.060)
−28.81***
(0.073)
603,169
II. Controls: I + age and pregnancy categories
January
3364.26
3348.85
February
3375.05
3364.83
March
3367.79
3347.18
April
3372.12
3342.53
N
−15.41***
(0.062)
−10.22***
(0.074)
−20.61***
(0.067)
−29.59***
(0.072)
627,661
3275.20
3263.49
3278.15
3273.69
3271.51
3257.47
3282.06
3253.20
−11.71***
(0.063)
−4.46***
(0.063)
−14.04***
(0.077)
−28.86***
(0.081)
596,162
III. Controls: II + mother’s education and partner dummy variables
January
3317.01
3304.15
February
3327.10
3317.88
March
3320.76
3302.77
April
3324.56
3298.24
N
Note: See Table 2.
591,593
−12.86***
(0.220)
−9.22***
(0.226)
−17.99***
(0.207)
−26.31***
(0.205)
3228.64
3216.15
3231.94
3227.64
3223.86
3211.35
3235.29
3207.26
561,864
−12.61***
(0.193)
−4.31***
(0.168)
−12.58***
(0.186)
−28.12***
(0.211)
Table 4: Differences in average birth weight (g) between 2002 and 2001 by
gender broken down by mother’s education
January
February
March
April
N
I. Controls: province dummy variables
Boys
Girls
Low
High
Low
High
−23.06*** −2.72***
−19.61*** 2.15***
(0.058)
(0.073)
(0.093)
(0.082)
−13.57*** −4.55***
−7.81*** −1.76 ***
(0.105)
(0.132)
(0.071)
(0.134)
−31.08*** −4.09***
−20.57*** −6.61***
(0.122)
(0.089)
(0.071)
(0.079)
−33.21*** −23.68***
−34.59*** −18.62***
(0.120)
(0.093)
(0.082)
(0.117)
392,993
225,007
374,852
212,226
II. Controls: I + age and pregnancy categories + partner dummy variables
Boys
Girls
Low
High
Low
High
January
−20.37*** 0.796***
−19.16*** −0.445
(0.262)
(0.244)
(0.253)
(0.279)
February
−13.67*** −1.94***
−5.61*** −2.66***
(0.266)
(0.262)
(0.212)
(0.241)
March
−28.28*** −0.142
−16.50*** −5.99***
(0.273)
(0.256)
(0.230)
(0.238)
April
−29.88*** −20.70***
−35.84*** −15.14***
(0.264)
(0.235)
(0.282)
(0.251)
N
Note: See Table 2.
377,109
214,484
359,804
202,060
Table 5: Regressions of Birth Weight on Economic Cycle in the Month of Birth
Cycle in the Month of Birth
Female
(1)
(2)
(3)
(4)
(5)
(6)
81.43**
(36.41)
115.66***
(25.86)
73.67*
(38.06)
111.18***
(26.57)
68.57*
(38.52)
99.35***
(25.85)
−102.92*** −102.92*** −103.08*** −103.08*** −103.44*** −103.44***
(0.757)
(0.757)
(0.791)
(0.791)
(0.826)
(0.826)
Time and region controls
Month of birth fixed effects?
Province of birth fixed effects?
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Year of birth fixed effects?
Linear time trend
Yes
--
No
−1.14***
(0.109)
Yes
--
No
−1.22***
(0.113)
Yes
--
No
−1.24***
(0.111)
Mother’s and pregnancy controls
Mother’s age categories?
Parity categories?
No
No
No
No
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Mother’s High School
--
--
--
--
Mother’s Partner Status
--
--
--
--
24.02***
(3.44)
58.43***
(1.72)
24.01***
(3.45)
58.44***
(1.72)
Mother’s High School × Mother’s
Partner Status
No
No
No
No
−17.18***
(3.36)
−17.18***
(3.36)
N
1,803,585
1,782,311
1,689,913
Note: All regressions include a constant term. Robust standard errors clustered at the “month-year” of birth level are reported in parentheses.
Month of birth fixed effects: 11 dummy variables; Province of birth fixed effects: 24 dummy variables; Year of birth fixed effects: 2 year
dummy variables; Linear time trend = 1…, 36; Mother’s age categories: 6 dummy variables (15-19, 20-24, 25-29, 30-34, 35-39, 40-44);
Parity categories: 3 dummy variables (1st pregnancy, 2nd pregnancy, 3rd pregnancy); Mother’s High School: 1 if mother has high-school or
above, 0 otherwise; Mother’s Partner Status: 1 if mother is living with a partner, 0 otherwise.
*** p-value < 0.01, ** p-value < 0.05, * p-value < 0.1
Table 6: Regressions of Birth Weight on Economic Cycle in the Month of Birth by Gender
Boys
Girls
(1)
(2)
(3)
(4)
83.93*
(44.60)
103.57***
(26.10)
52.73
(35.99)
94.89***
(27.60)
Time and region controls
Month fixed effects?
Province fixed effects?
Year fixed effects?
Linear Time Trend
Yes
Yes
Yes
--
Yes
Yes
No
−1.24***
(0.119)
Yes
Yes
Yes
--
Yes
Yes
No
−1.24***
(0.117)
Mother’s and pregnancy controls
Mother’s age categories?
Parity categories?
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
26.65***
(4.68)
58.65***
(2.38)
−13.70***
(4.53)
26.65***
(4.68)
58.66***
(2.38)
−13.70***
(4.54)
Cycle in the Month of Birth
Mother’s High School
Mother’s Partner Status
Mother’s High School × Mother’s Partner Status
N
Note: See Table 5.
866,812
21.20***
21.18***
(3.71)
(3.72)
58.28***
58.28***
(2.37)
(2.37)
−20.86*** −20.86 ***
(3.86)
(3.86)
823,101
Table 7: Regressions of Birth Weight on Economic Cycle in the Month of Birth broken down by mother’s
education
Mother’s Education < High School
Mother’s Education >= HS
(1)
(2)
(3)
(4)
(5)
(6)
Cycle in the Month of Birth
250.69***
(31.15)
246.22***
(33.64)
142.45***
(28.34)
Female
−98.53***
(1.06)
−98.96***
(1.13)
−98.99*** −110.58*** −110.98*** −110.94***
(1.13)
(1.39)
(1.43)
(1.42)
Time and region controls
Month fixed effects?
Province fixed effects?
Linear Time Trend
No
No
--
Yes
Yes
--
Yes
Yes
−1.29***
(0.124)
No
No
--
Yes
Yes
--
Yes
Yes
−1.15***
(0.112)
Mother’s and pregnancy controls
Mother’s age categories?
Parity categories?
No
No
Yes
Yes
Yes
Yes
No
No
Yes
Yes
Yes
Yes
Mother living with a partner
--
56.22***
(1.78)
55.99***
(1.78)
--
46.60***
(2.68)
46.53***
(2.70)
1,099,356
1,059,925
1,059,925
653,899
629,988
629,988
N
Note: See Table 5.
122.09***
(30.58)
112.32***
(30.81)
24.92
(24.81)
Table 8: Regressions of Birth Weight on Economic Cycles during Trimesters of Pregnancy
(1)
(2)
(3)
(4)
(5)
(6)
133.76***
(21.20)
--
--
--
--
--
Cycle 2nd Trimester of Pregnancy
203.28***
(41.11)
--
--
--
--
202.12***
(14.50)
--
--
Cycle 1st Trimester of Pregnancy
245.39***
(19.09)
--
182.27***
(19.36)
204.36***
(21.48)
Cycle 3rd Trimester of Pregnancy
Female
−103.45*** −103.44*** −103.45*** −103.45*** −103.46*** −103.45***
(0.825)
(0.825)
(0.825)
(0.825)
(0.826)
(0.825)
Time and region controls
Month fixed effects?
Province fixed effects?
Year fixed effects?
Linear Time Trend
Yes
Yes
Yes
--
Yes
Yes
No
−0.988**
(0.111)
Yes
Yes
Yes
--
Yes
Yes
No
−0.491***
(0.098)
Yes
Yes
Yes
--
Yes
Yes
No
−0.362**
(0.153)
Mother’s and pregnancy controls
Mother’s age categories?
Parity categories?
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
24.03***
(3.44)
58.36***
(1.72)
24.04***
(3.44)
58.36***
(1.72)
24.11***
(3.44)
58.35***
(1.72)
24.10***
(3.44)
58.33***
(1.72)
24.12***
(3.44)
58.41***
(1.72)
24.09***
(3.44)
58.52***
(1.72)
−17.19***
(3.36)
−17.19***
(3.36)
−17.26***
(3.36)
−17.24***
(3.36)
−17.26***
(3.36)
−17.28***
(3.36)
Mother’s High School
Mother’s Partner Status
Mother’s High School × Mother’s
Partner Status
N
Note: See Table 5.
1,689,913
1,689,913
1,689,913
Table 9: Regressions of Birth Weight on Economic Cycles during Trimesters of Pregnancy
(1)
(2)
(3)
(4)
(5)
(6)
128.70***
(32.90)
65.64
(53.82)
107.93***
(33.39)
91.43**
(36.01)
43.47
(53.86)
126.26***
(33.27)
119.74***
(33.15)
62.36
(54.41)
119.58***
(33.73)
85.84**
(35.91)
42.12
(54.38)
136.28***
(33.46)
113.28***
(33.87)
51.47
(54.04)
126.23***
(33.80)
79.31**
(36.18)
29.46
(55.55)
144.10***
(34.25)
Sum of Coefficients on Cycle 302.27***
(22.98)
261.16***
(17.67)
301.69***
(23.63)
264.24***
(17.59)
290.99***
(23.32)
252.87***
(17.61)
Cycle 3rd Trimester of Pregnancy
Cycle 2nd Trimester of Pregnancy
Cycle 1st Trimester of Pregnancy
Female
−102.93*** −102.93*** −103.10*** −103.10*** −103.46*** −103.46***
(0.756)
(0.756)
(0.789)
(0.789)
(0.825)
(0.826)
Time and region controls
Month fixed effects?
Province fixed effects?
Year fixed effects?
Linear Time Trend
Yes
Yes
Yes
--
Yes
Yes
No
−0.183**
(0.090)
Yes
Yes
Yes
--
Yes
Yes
No
−0.224**
(0.091)
Yes
Yes
Yes
--
Yes
Yes
No
−0.263***
(0.093)
Mother’s and pregnancy controls
Mother’s age categories?
Parity categories?
No
No
No
No
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Mother’s High School
--
--
--
--
Mother’s Partner Status
--
--
--
--
24.11***
(3.44)
58.34***
(1.72)
24.12***
(3.44)
58.34***
(1.72)
Mother’s High School × Mother’s
Partner Status
No
No
No
No
−17.26***
(3.36)
−17.26***
(3.36)
N
Note: See Table 5.
1,803,585
1,782,311
1,689,913
Table 10: Regressions of Birth Weight on Economic Cycles during Trimesters of Pregnancy by Gender
Boys
Girls
(1)
(2)
(3)
(4)
144.97***
(28.52)
28.50
(41.89)
148.65***
(30.67)
99.28***
(34.18)
−1.08
(50.28)
172.59***
(33.38)
79.67
(58.26)
76.97
(103.46)
101.82
(67.16)
57.63
(58.02)
62.66
(101.79)
113.47*
(66.08)
Sum of Coefficients on Cycle 322.12***
(23.69)
270.80***
(19.71)
258.46***
(33.05)
233.75***
(26.76)
Cycle 3rd Trimester of Pregnancy
Cycle 2nd Trimester of Pregnancy
Cycle 1st Trimester of Pregnancy
Time and region controls
Month fixed effects?
Province fixed effects?
Year fixed effects?
Linear Time Trend
Yes
Yes
Yes
--
Yes
Yes
No
−0.197**
(0.099)
Yes
Yes
Yes
--
Yes
Yes
No
−0.332**
(0.140)
Mother’s and pregnancy controls
Mother’s age categories?
Parity categories?
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
26.73***
(4.67)
58.57***
(2.38)
−13.76***
(4.53)
26.74***
(4.67)
58.56***
(2.38)
−13.76***
(4.52)
21.31***
(3.71)
58.19***
(2.38)
−20.97***
(3.86)
21.31***
(3.71)
58.20***
(2.38)
−20.97***
(3.86)
Mother’s High School
Mother’s Partner Status
Mother’s High School × Mother’s Partner Status
N
Note: See Table 5.
866,812
823,101
Table 11: Regressions of Birth Weight on Economic Cycles during Trimesters of Pregnancy by
Mother’s Education
Mother’s Education <
High School
(1)
(2)
Mother’s Education ≥
High School
(3)
(4)
182.51***
(45.73)
25.84
(68.76)
141.12***
39.94
143.48***
(48.48)
3.00
(72.62)
159.95***
(42.98)
6.93
(41.64)
86.48
(60.38)
110.02**
(43.96)
−24.35
(36.64)
63.00
(59.05)
128.64***
(41.58)
Sum of Coefficients on Cycle 349.47***
31.57
306.43***
(20.89)
203.43***
(30.84)
167.29***
(27.95)
Cycle 3rd Trimester of Pregnancy
Cycle 2nd Trimester of Pregnancy
Cycle 1st Trimester of Pregnancy
Female
−99.01***
(1.13)
−99.01*** −110.97*** −110.97***
(1.13)
(1.42)
(1.42)
Time and region controls
Month fixed effects?
Province fixed effects?
Year fixed effects?
Linear Time Trend
Yes
Yes
Yes
--
Yes
Yes
No
−0.222*
(0.112)
Yes
Yes
Yes
--
Yes
Yes
No
−0.305**
(0.142)
Mother’s and pregnancy controls
Mother’s age categories?
Parity categories?
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
55.88***
(1.79)
55.89***
(1.78)
46.41***
(2.70)
46.40***
(2.70)
Mother’s Partner Status
N
Note: See Table 5.
1,059,925
629,988
Table 12: A simple calculation of the Future Income Loss due to Lower Birth
Weight, in PPP International Dollars of 2009.
Annual Income Growth
1%
3%
5%
Annual
2%
452
925
1981
Discount Factor
5%
175
328
643
8%
78
136
245
Note: Calculation assumes Δln(Wage)/Δln(BW) = 0.09 (lower bound from Black,
Devereux and Salvanes, 2007), Δln(BW) = −0.0091 (mean birth weight in singletons
2002-03 vs. 2001), annual income in 2009 = $ 14,559 (GDP per capita, PPP, 2009).
Individuals earn income between age 22 and 65 (for an individual born in 2002 this
represents the period 2024-2067). The discounted income loss is calculated as the
difference between income with and without birth weight loss, where the gap is
calculated using the estimates from Black, Devereux and Salvanes (2007) and above,
and the birth weight gap mentioned above (gap = 0.11*0.0091 ≅ 0.001). Income at year
t is Y t = 14559(1+g)(t-2009), the income loss in year t in dollars is Y t (1 – gap) and the
present value using discount δ is
2065
∑ (1 − δ )
t = 2024
t − 2009
Yt (1 − gap)
Table 13: Robustness checks with additional controls
(1)
(2)
(3)
(4)
(5)
(6)
111.19***
(36.95)
65.57
(62.53)
109.16***
(37.72)
79.22*
(40.78)
44.82
(63.17)
126.02***
(38.14)
142.72***
(40.68)
22.62
(66.96)
131.40***
(40.09)
107.72**
(43.82)
6.65
(67.05)
147.60***
(39.51)
110.53***
(37.42)
72.30
(62.78)
105.51***
(38.21)
84.64**
(40.29)
55.28
(62.70)
119.21***
(38.32)
Sum of Coefficients on Cycle 285.93***
(25.28)
250.06***
(18.99)
296.73***
(26.46)
261.97***
(20.28)
288.34***
(25.33)
259.13***
(19.54)
Yes
Yes
Yes
--
Yes
Yes
No
−0.217**
(0.102)
Yes
Yes
Yes
--
Yes
Yes
No
−0.214***
(0.107)
Yes
Yes
Yes
--
Yes
Yes
No
−0.128
(0.103)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
No
No
No
No
No
No
Yes
Yes
No
No
No
No
No
No
Yes
Yes
Cycle 3rd Trimester of Pregnancy
Cycle 2nd Trimester of Pregnancy
Cycle 1st Trimester of Pregnancy
Time and region controls
Month fixed effects?
Province fixed effects?
Year fixed effects?
Linear Time Trend
Mother’s and pregnancy controls
Mother’s age categories?
Parity categories?
Mother’s Education Status
Mother’s Partner Status
Mother’s High School × Mother’s
Partner Status
New controls
Doctor aiding the delivery of the
baby indicator?
Mother’s health insurance coverage
indicators?
Place of birth (public hospital,
private hospital, home, street)
indicators?
N
Note: See Table 5.
1,737,331
1,541,482
1,754,872
Table 14: Regressions of Birth Weight on Indigence and Poverty Rates
Indigence 3rd Trimester
Indigence 2nd Trimester
Indigence 1st Trimester
(1)
(2)
(3)
(4)
(5)
−0.961
(0.892)
−0.629
(1.515)
0.615
(1.017)
--
−1.316***
(0.381)
--
---
−0.898**
(0.403)
--
Poverty 3rd Trimester
--
Poverty 2nd Trimester
--
Poverty 1st Trimester
--
---
0.283
(0.337)
--
--
0.158
(0.833)
−1.645
(1.285)
1.265
(0.791)
--
−0.862**
(0.392)
--
−0.528
(0.438)
--
--
0.356
(0.238)
--
--
Joint Sign. F-test
p-value F-test
6.10***
0.003
2.16
0.120
8.95***
0.001
2.43
0.111
14.72***
0.000
N
Period Covered
1,125,353
Jun01/
May03
1,125,353
Jun01/
May03
1,125,353
Jun01/
May03
1,125,353
Jun01/
May03
1,125,353
Jun01/
May03
Note: Robust standard errors clustered at the “month-year” level. Controls: all used in column (5) in Table 5
(except business cycle indicators).
*** p-value < 0.01, ** p-value < 0.05,* p-value < 0.1.